As time passes, AI's technology has been developed to provide us with more and more comforts and amenities in our daily lives. Continuing in this direction, AI Research's team Facebook in partnership with Carnegie Mellon University (CMU), has developed an AI that can win professional poker players.
The new AI system, called Pluribus, was announced through one press release, managed to beat 12 from the best poker players in the popular six-player no-limit Texas Hold'em.
Pluribus played against 12 players in two batches. In the first bout AI played against 5 people and in the second was five AIs and one man. AI bot was able to earn an average of up to $ 1.000 per hour - a figure that is considered very impressive for an AI.
Thousands of players participated in the fight over a two-day period.
The list of poker players includes Jimmy Chou, Seth Davies, Michael Gagliano, Anthony Gregg, Dong Kim, Jason Les, Linus Loeliger, Daniel McAuley, Nick Pietrangelo, Sean Ruane, Trevor Savage and Jake Toole.
It is believed that Pluribus was able to win it game, applying the Nash equilibrium theory that involves calculating equal winning probabilities in a game, as is done in the stone-scissors-paper toy.
Pluribus comes after a series of other experiments such as Libratus AI, which was able to defeat a player when playing with an opponent. The new discovery indicates that its technology artificial intelligence can achieve superhuman skills in new challenges.
How was Pluribus trained?
In order to beat the top poker players, Pluribus was initially trained to play with different versions of himself to learn to play through his trials and failures.
After a while, he was able to predict some of his opponent's moves, but not throughout the game, by applying a search in real time.
It took only eight days to train the AI while 64 core servers were used, less than 512 GB of RAM and not at all GPU. Still, this AI is more cost-effective than others.
It may not seem to be a great achievement, but poker is one of the hardest paper games, as it is about bluffing and predicting the opponent's move, where sometimes even people fail. This AI feature is very important.